"(\\d+)\s*год\\b": "\\1", as originally suggested by @noviluni, works for the new test cases, but fails for existing ones.

Because I do not speak Russian, I am not sure how to best address those other issues.

For example, according to tests, 1 год 2 месяца means 1 year and 2 months ago. Is that correct? If so, does 1 год mean 1 year ago or year 1? How can you tell?

Hi @Gallaecio

1 год 2 месяца means 1 year and 2 months ago. Is that correct?

Yes, you're right.

If so, does 1 год mean 1 year ago or year 1?

It means year 1.

I think I just overcomplicated the tests and most of the web pages have a simpler format.

How can you tell 1 год means year 1 but 1 год 2 месяца means 1 year and 2 months ago? How would you say 1 year ago in line with 1 год 2 месяца?

1 year ago - один год назад / 1 год назад
2001 years ago - 2001 год назад
year 2001 - 2001 год

In Russian, there would be one year(1 год), two years(2 года), three years(3 года), and four years(4 года), but five years(5 лет), six years(6 лет), seven years(7 лет), eight years(8 лет), and nine years(9 лет).

That is, to count the years, you need two words: год and лет.

In general,
2000 год - it's the year 2000
2000 лет назад - 2000 years ago
2001 год - it's the year 2001
2001 год назад - 2000 years ago

1 год 2 месяца means 1 year and 2 months ago. Is that correct?

Yes, you're right.

is because the word the месяц(month) can be used with the words either backward or forward in time.

If so, does 1 год mean 1 year ago or year 1?

It means year 1.

it's because for ago need a world назад.

OK, so I have gone with 2 regular expressions:

• Given a number of 4+ digits ending in 1 and followed by год and only by год, remove the “years” suffix so that it is treated as an absolute year rather than a number of years. I went with this after I wound that parse("1000") interprets 1000 as the year, but parse("999") gives no result.
• Given any number not ending in 1 and followed by год, remove the “years” suffix so that it is treated as an absolute year.

I have added a few more test cases that I hope are accurate. Please take a good look.

LGTM @Gallaecio @wRAR